sklearn.ensemble._weight_boosting.BaseWeightBoosting.fit
for iboost in range(self.n_estimators):
# Boosting step
sample_weight, estimator_weight, estimator_error = self._boost(
iboost,
X, y,
sample_weight,
random_state)
sample_weight_sum = np.sum(sample_weight)
if iboost < self.n_estimators - 1:
# Normalize
sample_weight /= sample_weight_sum
sklearn.ensemble._weight_boosting.AdaBoostClassifier._boost_discrete
estimator.fit(X, y, sample_weight=sample_weight)
sample_weight
Is wmi w_{mi}wm i, Fitting according to sample weight
# Instances incorrectly classified
incorrect = y_predict != y
# Error fraction
estimator_error = np.mean(
np.average(incorrect, weights=sample_weight, axis=0))
The misclassification rate is the sum of the weights of the misclassified samples: em = ∑ i = 1 N wmi I (G m (xi) ≠ yi) e_m=\sum_{i=1}^{N}w_{mi}I(G_m(x_i )\neq y_i)em=∑i=1Nwm iI(Gm(xi)=Yi)
If the effect of the weak learner is not even random, stop early.
Calculate G m G_mGmThe coefficient of α m = 1 2 log 1 − emem \alpha_m=\frac{1}{2}log\frac{1-e_m}{e_m}am=21logem1−em
SAMME
The algorithm considers multiple classifications, and multiplies it by the attenuation of learning_rate to achieve regularization.
w m i = e x p ( − y i α m − 1 G m − 1 ( x i ) ) w_{mi}=exp(-y_i \alpha_{m-1} G_{m-1}(x_i)) wm i=e x p ( - yiam−1Gm−1(xi))
试想 ,y × y ^ y \ times \ hat {y}Y×Y^, In the second classification, only when the two are different, it is 1
In fact, this writing also takes into account the multi-category situation. For multi-category, you only need to modify α m \alpha_mamThat is estimator_weight
the calculation in the code .